Dublin Tree Canopy Study March, 2017.

Michael Brennan, Gerald Mills and Tine Ningal

Dublin Tree Canopy Study

Final Report

Michael Brennan, Gerald Mills and Tine Ningal

28/03/2017

University College Dublin

School of Geography,

Newman Building, Belfield

Dublin 4

1

Contents

1. Introduction _________________________________________________________________ 2

2. Methods and results __________________________________________________________ 2

2.1 Deliverable 1: Assessment of total canopy cover and its geographical distribution ________________ 2

Urban canopy cover ___________________________________________________________________ 7

Rural canopy cover ___________________________________________________________________ 9

Tree canopy cover on OPW sites ________________________________________________________ 10

2.2 Deliverable 2: Estimation of the proportion of canopy cover in public ownership ________________ 12

Results ____________________________________________________________________________ 12

2.3 Deliverable 3: Comparison of Dublin canopy coverage with other major European cities: _________ 14

Results ____________________________________________________________________________ 14

2.4 Deliverable 4: Estimation of environmental services provided _______________________________ 19

2.4.1 Air quality _____________________________________________________________________ 19

2.4.2 Runoff ________________________________________________________________________ 23

3. Conclusions & Recommendations _______________________________________________ 28

4. References _________________________________________________________________ 30

Appendix 1. ___________________________________________________________________ 33

Estimating canopy cover: Comparison with other methods ____________________________ 33

Appendix 2. ___________________________________________________________________ 40

The databases generated for the DTCS _____________________________________________ 40

2

1. Introduction This report describes the work carried out as part of the Dublin Tree Canopy Study (DTCS). The DTCS

undertook a complete aerial survey of the tree canopy cover for the Greater Dublin Region within

the lands managed by the four Dublin local authorities (Fingal (FCC), Dublin City (DCC), Dun

Laoghaire-Rathdown (DLR) and South Dublin (SDCC)), and the Office of Public Works (OPW). The

DTCS has five deliverables as described in the project methodology (Brennan et al., 2015), these

being:

1. Assessment of the total canopy coverage and its geographical distribution

2. Estimation of the proportion of cover that is in public ownership

3. Comparison of Dublin canopy coverage with other major European cities

4. Estimation of the environmental services provided by the current cover

5. Recommendations for tree canopy cover

This report proceeds with an overview of the methods used to achieve Deliverables 1-4, with

Deliverable 5 being addressed in the conclusion.

2. Methods and results

2.1 Deliverable 1: Assessment of total canopy cover and its geographical distribution

To achieve Deliverable 1 several methodologies were employed. Firstly, satellite imagery for the

entire study area was downloaded using the SAS Planet software. This imagery was geo-rectified

(Pollefeys et al., 1999) and organised into larger image tiles for processing within the ArcGIS

platform. Supervised classification tools within ArcGIS were used in an attempt to classify the

imagery. However, despite initially promising results, these tools were unable to routinely

distinguish tree canopy from other vegetated land cover classes (e.g. grass, shrubs, etc.), and

significant misclassification of non-vegetated land covers (e.g. roads) occurred in shadowed areas.

In light of the difficulties encountered with image classification, another remote sensing approach

was implemented based on spatial sampling. This is a two-step process: first, a large number of

sample points are generated and superimposed upon imagery of the study area, and; second, the

land-cover below each point is visually classified into as one of 13 categories. This is a well-tested

methodology that is similar to the approach adopted by the i-Tree software1 (Nowak et al., 1996).

Table 1 describes the land cover classes used in the DTCS.

Sampling methodologies have been routinely employed in other jurisdictions to establish baseline

tree canopy cover and monitor changes over time (Heynen et al., 2006, Berland, 2012, Greater

London Authority, 2015). Though the number of points used varies between studies, guidance

provided by the London Tree Officers Association2 suggests at least 3,000 points be used per London

borough (15-150 km2), resulting in a sampling density of 0.2-2 points per hectare (ha). Given that the

DTCS study area consists of ~92,600 ha (926km2), this suggests that 18,521- 185,210 points be used.

1 https://canopy.itreetools.org/ 2 http://www.ltoa.org.uk/resources/open-source-canopy-cover-audit-oscca

3

Table 1. Land cover classifications used in the DTCS

Land cover label Description

1. Tree cover - deciduous Deciduous trees, both single trees and woodland

2. Tree cover - coniferous Coniferous trees, both single trees and woodland

3. Shrub cover Ambiguous woody plants/Unmanaged and overgrown land

4. Herbaceous vegetation / Grassland

All types of grass covered land, including grazing land, amenity and domestic lawns, etc.

5. Cultivated and managed / Cropland

Includes cereals, vegetables, fruits and lands under glass/polythene

6. Regularly flooded / wetland Includes marshes, bogland with surface water, flooded quarries, plant choked ponds, etc.

7. Urban / built up - buildings All built structures

8. Urban / built up - roads All road surfaces, both asphalt and concrete

9. Urban / built up - other impervious surfaces

Includes paved areas, footpaths, (non-grass) sports pitches, driveways

10. Snow and ice Snow and ice

11. Barren Bare rock and soil, can include dirt tracks

12. Open water All bodies of open water.

13. Not Sure Uncertain land cover

However, preliminary results from a range of test locations established that using points within this

this density range, even at maximum, were not appropriate in all areas of the study area. In urban

areas, considerably more points were found to be necessary to capture an accurate estimate of tree

canopy due to the highly heterogeneous nature of the urban land cover. In contrast to rural areas,

having a generally more homogeneous land cover, required a much lower level of sampling to

estimate tree canopy cover accurately. The study area was divided into urban and rural areas using

the 2012 CORINE land cover dataset (Corine Land Cover, 2000, EPA, 2012). In the CORINE scheme,

‘urban’ corresponds closely to the artificial landcover category.3

To ensure geographical cover a systematic random sampling method was employed. In the urban

areas of the study area 440,397 points were sampled at a rate of once each 30 m (resulting in

approximately 12 points per hectare). In the rural areas, a sampling rate of 100 m was employed,

resulting in 25,000 sample points (0.44 points per hectare). Altogether, 464,386 sample points were

categorised using the scheme shown in Table 1. Once classification of all points in both rural and

urban areas was completed the points were aggregated into grid cells, with the proportion of canopy

cover in each cell calculated as:

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑖𝑛𝑡𝑠 𝑐𝑙𝑎𝑠𝑠𝑖𝑓𝑖𝑒𝑑 𝑎𝑠 𝑡𝑟𝑒𝑒

𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑖𝑛𝑡𝑠= 𝑃𝑟𝑜𝑝𝑜𝑟𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡𝑟𝑒𝑒 𝑐𝑜𝑣𝑒𝑟

Owing to the different sampling densities employed in rural and urban landscapes, urban cells have

a cell size of 250m, and rural areas have a cell size of 1km – in other words the spatial detail for

urban areas is 16 times more precise. The area of canopy cover in each cell was then calculated as:

𝑃𝑟𝑜𝑝𝑜𝑟𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡𝑟𝑒𝑒 𝑐𝑜𝑣𝑒𝑟 ×𝐶𝑒𝑙𝑙 𝑎𝑟𝑒𝑎 = 𝑇𝑟𝑒𝑒 𝑐𝑎𝑛𝑜𝑝𝑦 𝑎𝑟𝑒𝑎

The canopy area values for each cell were then summed for the entire study area, urban areas, rural

areas and each of the four Dublin local authority areas.

3 http://image2000.jrc.ec.europa.eu/reports/technical_guide.pdf

4

The canopy assessment was completed at the end of January 2016. The full results expand on

findings presented at the Urban Forest event held in Wood Quay4, with tree canopy unevenly

distributed throughout the study area (Figure 1). Total canopy is estimated to be 9284 ha, covering

approximately 10% of the entire study area.

Figure 1. Total tree canopy cover assessment within the study area. A cell size of 250m is used in

urban areas and 1000m in rural areas.

4 https://www.youtube.com/watch?v=rAC-YJK39X0&index=3&list=PLHKVjBSDqMB7ZJCrB1i6qjRLQQ5qrGm23

5

Table 2. Canopy cover in each of the local authorities and for the study area.

Local Authority

Area (ha)

Canopy (ha)

Percent canopy

Percent urban

DCC 11,772 1,197 10.2 97.1

DLR 12,660 2,398 18.9 59.8

Fingal 45,806 2,996 6.5 25.7

SDCC 22,350 3872 17.3 42.2

Total 92,588 9,284 10.0 43.4

When total canopy cover is examined by local authority area (Table 2) there are significant

differences. Of the four LAs, Fingal has the lowest cover and DLR the highest. These numbers reflect

the very different landuses within the LAs, especially the division between urban and rural uses (as

defined by the Corine landcover database). Figures 2-5 shows the tree cover in each of the LAs.

Figure 2. Estimated tree canopy cover across Fingal, which has predominantly rural landcover (74%)

and the estimated overall tree canopy cover is estimated at 6.5%.

6

Figure 3. Estimated tree canopy cover across DCC, which has predominantly urban landcover (97%

and the estimated overall tree canopy cover is estimated at 10.2%.

Figure 4. Estimated tree canopy cover across SDCC, which is 42% urbanised. The estimated overall

total canopy cover is 17%

7

Figure 5. Estimated tree canopy cover across DLR, which is nearly 60% urbanised. The estimated

overall tree canopy cover is 18.9%.

Urban canopy cover

The mean canopy cover of urban cells is 10.71% however the distribution is positively skewed and

most cells have cover fractions of ≤10%. The total urban canopy area is estimated at 4,503.19 ha.

Differences are evident between the four Dublin local authority areas. By comparison with the study

area as a whole and the other partner jurisdictions (Tables 3 & 4 and Figure 6) DLR has a significantly

higher canopy cover (10-20% and above). An axis of low canopy cover stretches from the docklands,

through the inner city and extending through Crumlin to Tallaght. Other areas of low canopy cover

include the North Inner City, the various business parks proximate to the M50 motorway and the

Fingal towns of Rush, Lusk, Skerries and Balbriggan. Areas of high canopy cover (excluding DLR)

include a “green axis” beginning at the National Botanic Garden, stretching into Glasnevin, along

Griffith Avenue, into Clontarf and St. Anne’s Park. Other areas of high canopy cover include Howth,

Swords and Malahide in Fingal and Clondalkin in South Dublin.

8

Table 3. Proportion of urban grid cells in each canopy cover percentage class for each local authority

area.

Urban areas 0%-10% 10%-20% 20%-30% 30%-40% 40%-50% ≥50%

DCC 0.63 0.23 0.09 0.03 0.01 0.00

DLR 0.36 0.32 0.19 0.09 0.03 0.02

Fingal 0.64 0.22 0.08 0.04 0.01 0.02

SDCC 0.66 0.21 0.07 0.04 0.02 0.00

Total 0.59 0.23 0.10 0.04 0.02 0.00

Table 4. Total urban tree canopy area, urban canopy cover as a percentage of urban area, mean

canopy cover per urban cell for each local authority area.

Urban Area (ha)

Urban Canopy (ha)

% canopy cover

Mean canopy per cell (%)

Mean canopy per cell (ha)

DCC 11,425.27 1,185.11 10.37 10.02 0.58

DLR 7,565.79 1,238.90 16.38 15.83 0.81

Fingal 11,754.11 1,151.41 9.80 9.04 0.39

SDCC 9,426.95 927.77 9.84 9.07 0.42

Total 40,172.12 4,503.19 11.21 4.30 0.24

Figure 6. Proportion of urban grid cells (y-axis) in each canopy cover percentage (x-axis) class for

each local authority area.

9

Table 4a provides a breakdown of land-cover within the urban parts of the four LAs. As part of the

point sampling process the underlying cover was classified into one of nine categories (Table 1). The

distinction between tree and shrub is based on height and shape of the canopy. This table is a useful

assessment of the total green cover (the last column) in the urbanised part of Dublin county, most of

which is grass.

Table 4a. Percent land-cover in the urban parts of the four local authority areas.

Local Authority

Tree Shrub Grass Impervious Road Buildings Green

DCC 10.37 6.18 25.69 26.50 9.23 20.28 42.24

DLR 16.38 9.98 28.12 19.27 8.54 15.04 54.48

Fingal 9.80 9.18 37.80 18.10 8.23 12.20 56.78

SDCC 9.84 7.68 33.87 21.76 9.12 14.37 51.39

Rural canopy cover The majority of rural areas in the study area have relatively low tree canopy cover (0-10%), with the mean canopy cover being lower than that for urban area (8.35%). Total canopy area in rural canopy area is estimated at 4780.79 ha for the entire rural area. DLR, and SDCC are significantly different compared to both the study area as a whole and the other partner jurisdictions (Table 5 & 6 and Figure 7) as both include forested stands the Dublin Mountains, which have complete canopy cover. Table 5. Proportion of rural grid cells in each canopy cover percentage class for each local authority

area.

Rural areas 0%-10% 10%-20% 20%-30% 30%-40% 40%-50% ≥50%

DCC 0.83 0.11 0.06 0.00 0.00 0.00

DLR 0.40 0.20 0.15 0.12 0.10 0.04

Fingal 0.83 0.10 0.03 0.02 0.01 0.01

SDCC 0.61 0.17 0.08 0.03 0.04 0.06

Total 0.74 0.13 0.05 0.03 0.02 0.01

Table 6. Total rural tree canopy area, rural canopy cover as a percentage of rural area, mean canopy

cover per rural cell for each local authority area.

Rural Area (ha)

Rural Canopy (ha)

% canopy cover

Mean canopy per cell (%)

Mean canopy per cell (ha)

DCC 346.78 11.99 3.46 2.77 0.26

DLR 5,094.25 1,158.95 22.75 16.15 10.08

Fingal 34,052.25 1,844.47 5.42 5.91 3.43

SDCC 12,923.06 2,944.46 22.78 13.56 8.02

Total 52,416.35 4,780.79 9.12 8.35 0.24

10

Figure 7. Proportion of rural grid cells (y-axis) in each canopy cover percentage (x-axis) class for each

local authority area

Tree canopy cover on OPW sites

The sites managed by the OPW represent a special case (Table 7). The small number of sites which

are of interest to the OPW allowed the direct digitisation of the tree canopies within to be carried

out in ArcGIS. The OPW sites include: Arbour Hill Cemetery, Custom House, Dublin Castle, Farmleigh,

the Garden of Remembrance, Grangegorman Military Cemetery, the Iveagh Gardens, Leinster

House, the National Botanic Gardens, the National Gallery of Ireland, the National War Memorial

Gardens, St. Enda's Park, St. Stephen's Green and the Phoenix park.

Table 7. Digitisation status of each OPW site of interest

Site Canopy

Perimeter (m) Canopy

Area (ha) Site

Area (ha) Canopy

Cover (%)

Arbour Hill 1191.34 0.43 1.41 30.88

Custom House 576.13 0.14 1.78 8.11

Dublin Castle 643.47 0.13 6.21 2.09

Farmleigh 19724.67 9.90 32.19 30.75

Garden of Remembrance 401.52 0.23 0.54 42.71

Grangegorman 1765.27 0.56 2.12 26.58

Iveagh Gardens 3723.53 2.03 3.32 61.30

Leinster House 734.70 0.27 1.93 13.90

National Botanic Gardens 14997.55 8.07 18.38 43.90

National Gallery of Ireland 0.00 0.00 1.13 0.00

Phoenix Park 10987.8 182.33 708.07 25.75

St. Enda’s Park 8237.38 7.60 16.69 45.55

Stephens Green 8230.44 5.22 9.33 55.90

War Memorial Gardens 14415.71 9.14 22.67 40.32

11

Tree canopy has been digitised for all OPW sites (Figure 8). Despite a large range in canopy cover,

OPW sites generally contained higher canopy cover than the average urban cell, with a mean canopy

cover of 30.92%.

Figure 8. The digitised canopy within the OPW sites.

12

2.2 Deliverable 2: Estimation of the proportion of canopy cover in public ownership

To identify which trees are located in public spaces it was necessary to build a GIS dataset of likely

public spaces. This dataset is being created from three primary geographic datasets:

1. the road network,

2. the boundaries for the parks and other open spaces and,

3. other public spaces, such as those identified in DCC’s Habitat Mapping Initiative (O’Riain et

al., 2006).

The creation of the public space dataset followed a series of steps. First, the public space around the

road network was estimated by creating a 5m buffer around the centreline of the road network.

Second, these spaces were merged with the park and open space coverages (boundaries were added

to the public space dataset using the Merge function of ArcGIS. Finally, public spaces recorded in

additional datasets (e.g. the Dublin City Habitat Mapping Initiative) were extracted and merged with

the others. In subsequent analysis, the area covered by this dataset (hereafter referred to as Public

Space dataset) was regarded as representing the publicly owned areas within the study area.

Calculation of the proportion of tree canopy in public ownership then proceeded on two parallel

tracks. Firstly, the Public Space dataset was used to locate all sample points (from Deliverable 1)

located in public space. This was achieved using the simple Select by Location5 function in ArcGIS.

Secondly, the Public Space dataset was used to calculate the proportion of public space in each

250m urban grid cell used previously in the canopy analysis. Thus, each grid cell then contained data

which detailed the proportion of that cell which was covered by tree canopy and the proportion of

that cell which consisted of public space. By combining these pieces of information, it was possible

to estimate the proportion of the cell that was covered by tree canopy in public ownership.

Results

The land in public ownership is estimated to be 22,351ha or 24% of the study area (Table 8). Stark

differences were apparent between urban and rural areas in terms of the proportion of land in

public ownership; over 40% of urban land was estimated to be in public ownership versus slightly

under 11% in rural areas. This is to be expected, but highlights that the policy options available to

the project partners should they seek to increase canopy cover will by necessity be very different in

urban versus rural areas.

5 http://desktop.arcgis.com/en/arcmap/10.3/map/working-with-layers/using-select-by-location.htm

13

Table 8. Proportion of the study area estimated to be in public ownership, for the entire study area

and broken down into urban and rural components.

Area breakdown Area (ha) % of total

Total Study area 92,588.47 100

Total Public Space 22,351.20 24.14

Total Private Space 70,237.27 75.86

Rural

Rural area 52,416.35 100

Rural Public Space 5,724.90 10.92

Rural Private Space 46,691.45 89.08

Urban

Urban area 40,172.12 100

Urban Public Space 16,626.30 41.39

Urban Private Space 23,545.82 58.61

The proportion of tree canopy in public ownership varied greatly between urban and rural areas and

between Dublin counties. Due to the relatively small amount of rural public land, in-depth analysis of

rural public tree canopy (which would have accounted for an even smaller land area) is not feasible

without undertaking more data gathering and analysis. Of the estimated 4,503 ha of tree canopy

cover in urban areas, less than half (1,969 ha) was calculated as being in public ownership (Table 9).

DCC differs in character compared to the other Dublin authorities with the majority (60%) of its

canopy located on publicly held lands.

Table 9. The amount of total urban tree canopy cover, and the amount of canopy on publicly and

privately owned land disaggregated to county level.

Urban Area (ha)

Urban Canopy (ha)

Public Canopy (ha)

Private Canopy (ha)

Public Canopy (%)

Private Canopy (%)

DCC 11,425.27 1,185.11 721.36 463.75 60.87 39.13

DLR 7,565.79 1,238.90 409.09 829.81 33.02 66.98

Fingal 11,754.11 1,151.41 479.94 671.47 41.68 58.32

SDCC 9,426.95 927.77 358.16 569.61 38.60 61.40

Total 40,172.12 4,503.19 1,968.55 2,534.64 43.71 56.29

14

2.3 Deliverable 3: Comparison of Dublin canopy coverage with other major European

cities:

As part of this deliverable, published estimates of canopy cover in a wide range of cities were

compiled through literature review. Additionally, queries were directed to the COST Action FP12046

European research group, which has a focus on urban forestry. The literature review highlighted that

the calculation of urban canopy cover has been conducted for many cities globally, using a variety of

methods. This variety of methods can lead to a range of estimates for total canopy cover the same

city (Heynen et al., 2006), and can be extremely time consuming for large study sites.

Given the difficulties encountered with image classification, as described above, a single coherent

methodology was required to compare the canopy cover of Dublin to a selection of cities including

other Irish urban centres, UK and European cities. With this in mind, the i-Tree canopy7 online

canopy classification tool was chosen as it was freely available and would be implemented in a

standardised manner across all chosen cities. i-Tree canopy operates in a similar way to the method

used in Deliverable 1, i.e. a large number of random points are placed over an area of interest and

visually categorised by the operator. The base image used for this system is GoogleEarth. The cities

chosen for comparison were European capitals, a selection of UK cities and the major Irish urban

centres. The comparison area for each city was defined by a 1km circular buffer around the city

centre point. Within each comparison area 6,000 randomly selected points were generated and

these were classified into the land-cover categories shown in Table 1. The canopy cover was

calculated as the percent of points identified as canopy within the city centre buffer.

Results

There are a great number of publications on urban greening; a simple search for potentially relevant

publications in Google Scholar yields over 20,000 results. Our focus on places where we could get

comparable canopy statistics is shown in Table 10. As expected, there is a wide range of values in the

published literature for urban canopy cover in different cities. This variation reflects both varying

canopy cover and definitions of what constitutes the urban area so that simple comparison between

cities is not very useful. For North American cities the median cover is 26.3%; for major cities the

highest cover (>50%) reported was for Atlanta and the lowest (12.4%) was for Miami. The canopy

cover values for European cities were generally lower: the median value was 15.25% with the

highest reported value (57.3%) for Stockholm and lowest (0.9%) for Athens.

In this context, the overall value for the Dublin study area (10%) would place it in the lower half of

European city values; DLR and SDCC would be placed in the upper half, Fingal in the lower half and

Dublin city close to the average.

6 http://www.cost.eu/COST_Actions/fps/Actions/FP1204 7 http://www.itreetools.org/canopy/

15

Table 10. Estimated tree canopy cover in cities around the world based on published results. Sources

are listed in the references.

Source City Canopy

cover (%) Source City

Canopy cover (%)

North America Europe (Donovan and Butry, 2011)

Portland, Oregon 26 (Fryer, 2014) Cardiff, UK 15.6

(Szantoi et al., 2012) Miami-Dade County, Florida

12.4 (Fryer, 2014) Urban areas in Wales

16.8

(Locke et al., 2013) Baltimore, Maryland 27.4 i-Tree Survey 2014

Glasgow, UK 15

(McPhearson et al., 2013)

New York, New York 21 Red Rose Forest Survey 2007

Manchester, UK 15.5

(Nowak et al., 2013) Toronto, Ontario 26.6 (Davies and Doick, 2015)

Edinburgh, UK 17

(Merry et al., 2014) Detroit, Michigan 19.6 – 30.8 (Davies and Doick, 2015)

Torbay, UK 12

(Merry et al., 2014) Atlanta, Georgia 50 – 53 (Tallis et al., 2011) London, UK 20

(Locke and Grove, 2016)

Washington, D.C. 31.8 (Strohbach and Haase, 2012)

Leipzig, Germany

19

(Coble and Walsh, 2012)

St. Louis, Missouri 26 (Casalegno, 2011) Amsterdam, Netherlands

4.1

(Halter, 2013) Austin, Texas 33 (Casalegno, 2011) Athens, Greece 0.9

(Lee et al., 2016) Sacramento, California

18.2 (Casalegno, 2011) Berlin, Germany 42

(Potyondy, 2011). Hibbing, Minnesota 18.03 (Casalegno, 2011) Bruxelles, Belgium

12

(Boeur, 2010) District of Sechelt, British Columbia

59 (Casalegno, 2011) Budapest, Hungary

2.5

(Jorgensen, 2011) Saint Paul, Minnesota 32.5 (Casalegno, 2011) Copenhagen, Denmark

28.5

(2011) City of Bellingham

Bellingham, Washington

40 (Casalegno, 2011) Helsinki, Finland 49.1

(Shields, 2007) Kirkland, Washington 40 (Casalegno, 2011) Lisbon, Portugal 8.7

(Walton et al., 2008) Syracuse, New York 21.4 (Casalegno, 2011) Madrid, Spain 6.8

(Richardson and Moskal, 2014)

Seattle, Washington 28.5 (Casalegno, 2011) Paris, France 10.8

(McGee et al., 2012) Winchester, Virginia 27 (Casalegno, 2011) Rome, Italy 3.5

(Clayton, 2016) Seattle, Washington 23 (Casalegno, 2011) Stockholm, Sweden

57.3

(Crowley, 2011) Redlands, California 25 (Casalegno, 2011) Warsaw, Poland 36.5

(Wilson and Lindsey, 2009)

Indianapolis (CPD), Indiana

17 (Casalegno, 2011) Vienna, Austria 4.9

(Schwarz et al., 2015) Sacramento, California

54.64 Dublin Canopy Study

(Schwarz et al., 2015) Los Angeles, California

17.61 Dublin 10.0

Australia-Asia DCC 10.2

(Jacobs et al., 2014) Urban areas in Australia

27 – 56 DLR 18.9

(Jim and Chen, 2008) Guangzhou, China 7 Fingal 6.5

(Rahman and Agarwal) Bhopal, Madhya Pradesh, India

10.29 SDCC 17.3

16

For the direct comparison cities, i-Tree canopy assessments were completed. In addition, i-Tree

canopy assessment of Dublin city centre has also been completed. Compared to the city centres of

the other completed European and UK cities, Dublin city centre ranks lowest in terms of tree canopy

cover (Table 11). Similarly, Dublin has a lower level of canopy cover compared to all assessed Irish

urban centres with the exception of Belfast.

.

Table 11. Percentage tree canopy cover in comparison cities.

City # Points Mean % tree cover

Dublin 135 4.50

Belfast 177 3.54

Cork 456 9.12

Derry 810 16.2

Galway 332 6.64

Limerick 530 10.6 Cardiff 467 15.6

Amsterdam 232 7.74

Berlin 205 6.84

Brussels 189 6.31

Copenhagen 342 11.4

Helsinki 455 15.2

Lisbon 381 12.7

Madrid 198 6.6 Oslo 278 9.27

The green infrastructure (GI) present in Irish urban centres was compared by Mills et al. (2015), see

Table 11 and Figures 9 & 10. This study highlighted the variability of GI and tree canopy cover among

Irish urban centres. Much of the differences can be explained by the size of the city and the land-use

geography. The city centres of Dublin and Belfast are dominated by commercial land-uses and have

the lowest green cover and the lowest tree canopy cover; what GI is present is in the form of street

trees and public parks. In other cities, the study area includes low-density residential areas, where

there are often large private gardens. This is clearest for Derry and Waterford, the smallest cities

included in this study but it can also be seen in Limerick, where commercial development is

concentrated to the south of the Shannon extending westward. However, it is important to point out

that there are a large number of residents living in city centres; in 2011 there were approximately

40,000 living in the Dublin study area and 22,000 in the Cork study area. If the GI and tree canopy

cover are expressed in terms of population, then the differences between cities becomes stark.

Dublin has the lowest ratio of GI and canopy per capita, significantly lower than the values for

Belfast (Table 12).

The results show that Dublin city centre has the lowest cover of all the European city centres

compared. Some of this variation can be explained by local landuse but it is consistent. This indicates

that there is a need for a policy on GI in Dublin’s city centre that both conserves the existing stock

and expands tree planting to improve the urban environment for visitors and the growing resident

population.

17

Table 12. A classification of the land-cover in Irish city centres based on 5000 randomly selected

points within a circular area (1 km radius) using the i-Tree Canopy software. The values in each cell

represent the percentage of points (and standard error) that correspond to that land-cover. The final

three rows are the land area (that is the study area excluding water), the percentage of land area

that is Green (that is, Tree or Shrub or Grass) and Impervious (Roof, Road and Other impervious).

Source: Mills et al., 2015.

Land-cover Belfast Cork Derry Dublin Galway Limerick Waterford

Tree 3.54±0.26 9.12±0.41 16.2±0.52 4.52±0.29 6.64±0.35 10.6±0.44 6.14±0.34

Shrub 0.86±0.13 0.98±0.14 4.74±0.30 1.24±0.16 4.78±0.30 8.10±0.39 7.10±0.36

Grass 4.20±0.28 21.1±0.58 14.8±0.50 3.68±0.27 14.6±0.50 11.6±0.45 14.7±0.50

Road 36.4±0.68 26.4±0.62 9.92±0.42 18.4±0.55 13.2±0.48 13.0±0.48 14.3±0.50

Roof 32.0±0.66 31.1±0.65 24.7±0.61 39.6±0.69 22.1±0.59 24.1±0.60 28.3±0.64

Other Impervious 17.7±0.54 6.06±0.34 11.9±0.46 28.5±0.64 18.0±0.54 18.9±0.55 13.7±0.49

Water 4.44±0.29 5.10±0.31 14.2±0.49 3.88±0.27 11.3±0.45 12.3±0.46 13.8±0.49

Other 0.78±0.12 0.10±0.04 3.52±0.26 0.28±0.07 9.46±0.41 1.42±0.17 1.88±0.19

Land area (ha) 300 300 269 302 279 275 271

Green (%) 8.67 31.23 37.04 9.47 28.74 30.74 28.48

Impervious (%) 86.78 63.62 48.22 86.74 58.87 56.81 57.38

Figure 9. The Dublin city centre area study area as defined in i-Tree Canopy; 5000 randomly selected

points within the study area were classified into land-cover types. The red symbols represent those

sample points that fall on a tree canopy. The green areas represent the estimated green cover based

on the sampled points that represent trees, shrubs and grass. Source: Mills et al., 2015.

18

Figure 10. The study area of each of the Irish cities; the legend is the same as that for Figure 9.

Source: Mills et al., 2015.

19

2.4 Deliverable 4: Estimation of environmental services provided

Trees provide a number of environmental services (such as biodiversity, air quality, and water

management) but evaluating the value of these is not a simple task as it depends on species, age,

health, etc. of the tree stock.

2.4.1 Air quality

The tools for assessing air quality (AQ) services are best developed and are based on canopy plan

area, which is an indicator of the total leaf area exposed to the atmosphere. Trees remove pollutants

actively (through exchanges via stomata) and passively (by slowing air movement and providing

settling surfaces), hence the greater the leaf area exposed, the larger the potential removal. The

actual amount removed however depends on the ambient air quality to which the trees are

exposed, which is a function of the concentration of pollution (such as micrograms per cubic metre,

g m-3). The concentration itself is a product of both rate of pollutant emission and the volume of

atmosphere into which it is mixed. Naturally, the concentration of pollutants varies greatly in space

and time, especially in cities where there is a multitude of emission sources and mixing volumes. The

estimates shown here are based on typical removal functions associated with a climate and emission

properties similar to Dublin and are employed in i-Tree software (Morani et al., 2014).

Air quality is generally evaluated in terms of common urban pollutants that affect human health.

These include:

• Nitrogen dioxide (NO2) and Nitrogen oxides (NOx) which are emitted by internal combustion

engines.

• Sulphur Dioxide (SO2), which is linked to coal burning and is also emitted by diesel engines.

• Carbon monoxide (CO), which is a by-product of any fossil fuel combustion.

• Particulate matter (PM) less than 10 mm (PM10) and less than 2.5 mm (PM2.5), which are

sourced from road dust, tyre fragments, fires and combustion engines.

• Ozone (O3) which is a secondary pollutant created when other primary pollutants (such as

NOx) are present and atmospheric conditions are favourable.

The health impacts of these pollutants are measured in terms of threshold values that represent

either extreme values (daily or hourly) and mean background values. For some there are a set

number of days where the threshold values cannot be exceeded. Given the nature of the pollutants,

these values differ for each pollutant type (Table 13). Observations at a variety of stations around

Dublin (Table 14) are used to assess compliance. In general air quality in Dublin is described as good

and in its latest report the EPA indicated that no EU limit threshold values were exceeded. However,

potential challenges to improved air quality include switching from solid fuel to cleaner fuels for

home heating and the potential for increased NO2 pollution associated with transport.

Carbon dioxide (CO2) is not included in the urban air pollutants as it is not linked directly to human

health. However, the human contribution to global concentrations of CO2 is a major driver of climate

change and mitigating carbon emissions is a national objective. As trees play an important role in the

carbon cycle, the potential of the urban forest to offset CO2 emissions is examined.

20

Table 13. Summary of the Air Quality Directive’s limit values for the protection of human health.

Abstracted from Table A1 in EPA (2015)

Pollutant Averaging period

Value Num. of allowed occurrences.

SO2 Hour Day

350 g m-3

125 g m-3

24 3

NO2 Hour Day

200 g m-3

40 g m-3

3

CO Max daily 8-hour mean

10 mg m-3

PM10 Day Year

50 g m-3

20 g m-3

35 0

PM2.5 year 25 g m-3 0

O3 Max daily 8-hour mean

120 g m-3 25

Table 14. Annual means (g m-3) for air selected pollutants at stations in Dublin (EPA, 2015).

Po

lluta

nt

Win

eta

vern

St.

Co

lera

ine

St.

Rat

hm

ine

s

Du

n L

aogh

aire

Bal

lyfe

rmo

t

Bla

nch

ard

sto

wn

St A

nn

e’s

Par

k

Swo

rds

Bal

bri

ggan

Talla

ght

Clo

nsk

eag

h

Fin

glas

Mar

ino

NO2 31 25 17 15 16 31 14 14 9

NOx 59 41 27 22 25 67 22 24 14

SO2 1 1 3 7 6

CO 0 0.4 0.5

O3 43 51 53

PM10 14 14 14 11 18 17 23 15

PM25 9 9 7 8

Table 15 shows the estimated pollutant removal based on the plan area of canopy cover. The

estimates are provided for each local authority area and their rural and urban parts. The removal

functions are listed in the table caption and represent the estimated potential contribution of the

urban forest to Dublin’s air quality. The actual values would depend on the ambient air quality and

would require an analysis of the tree stock and the local climate. The estimated pollution removal by

the total urban forest is 11.8 tCO, 65.0 tNO2, 502 tO3, 25.6 tPM2.5 and 142 tPM10, which is about

equally divided between urban and rural contributions.

We can place these values into context by estimating the total amount of pollutants in the urban

atmosphere from Table 14. For example, a reasonable spatial average for urban NO2 is 20 g m-3 so

that 65 tNO2 represents 3250 km3 of cleaned air or a layer of air over the city 8 km high. Of course,

average NO2 values are an outcome of fresh emissions and losses due to chemical reaction and

advection; we might expect that the atmospheric boundary layer over Dublin refreshes itself about

once an hour (based on a 5 ms-1 windspeed) so that, at best, the removal by the canopy in the study

21

area accounts for one day’s emissions over a year. We could estimate the impact as a function of

vehicle emissions; assuming a vehicle emits 17 kg yr-1 of NOx and if 50-70% is converted to NO2 then

65 tNO2 is the equivalent of about 5000-7500 vehicles.

The carbon functions of trees can be assessed in terms of sequestration (CO2sq) and storage (CO2st),

which measures annual net carbon captured by leaves and stored in the woody skeleton,

respectively. The overall values are 103,702 tCO2sq and 1,687,549 tCO2st which are about evenly split

between urban and rural contributions. While the urban atmosphere is enriched with carbon, it is

not a local health issue. Instead the question: is to what extent the urban forest can offset CO2

emissions. If we assume that a vehicle emits about 250 gCO2 km-1 or 3.75 tCO2 yr-1 then the urban

forest can offset annual emissions from just 30 vehicles; moreover, the total CO2 in storage is worth

450 vehicles. The estimated emission within the DCC area was 5 million tCO2 in 2010 (DCC 2012): it

would take the entire urban forest in County Dublin about 5 years to sequester this quantity and

about 400 years for the DCC forest to do the same. So, while the trees provide a valuable function,

their air quality value should not be overstated. Much of the Dublin tree stock is deciduous so that

during winter the capacity of the tree stock to ‘clean’ the air is greatly reduced. However increasing

tree canopy cover as part of an air pollution strategy that manages emissions can improve air

quality.

The contributions of the tree canopy in each local authority can be evaluated simply as the

proportion of the forest cover in each local authority. Of course, there are significant differences

between each local authority in terms of canopy cover in rural and urban areas; DCC has by far the

smallest area classified as rural and perhaps the smallest capacity to increasing canopy cover (Table

5). If we restrict ourselves to just the urbanised part of the study area, DLR is exceptional as it has

the largest canopy percent (16%) and has nearly 1.5 times the canopy cover per unit area compared

to the other local authorities. However, much of this canopy is on private land (67%) which means

that the management of much of the GI relies on the goodwill of the property owner. By

comparison, DCC has the largest canopy in public space both in actual area and percent (Table 9).

22

Table 15. Area of tree canopy cover (hectares) in the study area decomposed into Local Authority (LA) and urban and rural landcover. The potential air quality benefits are shown as pollutant removal based on i-Tree Canopy software. Benefits are estimated based on tree canopy area cover using the following multipliers in g m2 yr-1: Carbon monoxide (CO) 0.127; Nitrogen Dioxide (NO2) 0.700; Ozone (O3) 5.404; Sulphur Dioxide (SO2) 0.344; Particulate matter less than 2.5 μm (PM2.5) 0.276 and less than 10 μm (PM10) 1.534; Carbon dioxide sequestered (CO2sq) 1,117 and; Carbon dioxide stored (CO2st) 18,177. The values in the table are in kg with the exception of CO2, which are in metric tonnes (1000 kg).

LA Area CO NO2 O3 SO2 PM2.5 PM10 CO2sq CO2st

Urban

DCC 1,185 1,505 8,295 64,037 4,076 3,271 18,178 13,236 215,397

DLR 1,239 1,574 8,673 66,956 4,262 3,420 19,006 13,840 225,213

Fingal 1,151 1,462 8,060 62,222 3,961 3,178 17,663 12,861 209,292

SDCC 928 1,178 6,494 50,137 3,192 2,561 14,232 10,363 168,641

Total 4,503 5,719 31,522 243,352 15,491 12,429 69,079 50,301 818,545

Rural

DCC 12 15 84 648 41 33 184 134 2,179

DLR 1,159 1,472 8,113 62,630 3,987 3,199 17,778 12,945 210,662

Fingal 1,844 2,342 12,911 99,675 6,345 5,091 28,294 20,603 335,269

SDCC 1,765 2,242 12,358 95,401 6,073 4,872 27,081 19,719 320,893

Total 4,781 6,072 33,466 258,354 16,446 13,195 73,337 53,401 869,004

Study area

Total 9,284 11,791 64,988 501,706 31,937 25,624 142,416 103,702 1,687,549

23

2.4.2 Runoff

That urban forests have the ability to regulate storm water flows is well known (Xiao and

McPherson, 2002). Urban trees perform this function by facilitating the infiltration of water into the

soil and evapotranspiration (Figure 11), as well as regulating the amount of throughfall reaching the

ground via direct rainfall interception (Cappiella et al., 2005). These processes reduce the flow rate

of urban runoff and shift the runoff concentration time by temporarily storing rain water on the

canopy surfaces (Sanders, 1986).

Figure 11. Processes provided by urban trees which regulate the runoff of precipitation (Levia and

Frost, 2006, Inkiläinen et al., 2013).

Decreasing the amount of storm runoff reduces the risk of flooding, reduces surface pollutant

washoff, and lowers the amount of pollutant in the runoff. Thus, by reducing urban runoff urban

trees ultimately reduce the cost of managing urban runoff and storm water treatment (Xiao et al.,

1998, Xiao and McPherson, 2002).

There have been a large number of studies exploring the mechanisms and extent to which urban

trees and forests regulate storm water flows (Sanders, 1986, Xiao et al., 1998, Xiao and McPherson,

2002, Link et al., 2004, Asadian and Weiler, 2009, Xiao and McPherson, 2011, Farrugia et al., 2013,

Inkiläinen et al., 2013) with a large range, from 10% to over 60% (Asadian and Weiler, 2009), in the

amount of rainfall being intercepted. How much rainfall is intercepted by urban trees depends on

the character of the rainfall event, the type of trees and their canopy extent, local climate and the

surrounding urban landscape. More intense and longer duration rainfall events have been

highlighted as important factors determining rainfall interception efficiency. Similarly, different tree

species can intercept different amounts of rainfall before becoming saturated, with conifers tending

to intercept a higher proportion of rainfall compared to broadleaves (Llorens and Domingo, 2007).

As with air quality, seasonal factors are important, with broadleaf trees losing much of their

interception capabilities in the winter. Finally, the height, composition and spacing of urban

buildings can enhance or reduce the delivery of rainfall to urban trees depending on their exact

characteristics.

24

The data required to quantify the amount of rainfall intercepted by an urban forest is extensive (Xiao and McPherson, 2011), and beyond the scope of this study. It requires not only knowledge of canopy area, but also leaf area index (LAI), tree height, tree diameter, stem-flow rate, canopy drip rate, and climatic variable to estimate evaporation (Xiao and McPherson, 2002). Some of this data (e.g. tree height, diameter and generalised LAI) could be gathered from existing sources, for example from street tree inventories8. That being said, studies have identified canopy area as being one of the key variables that explains the variations in rainwater between areas (Inkiläinen et al., 2013). However, the potential for the urban forest to regulate storm water flows should not be overstated. For example, Inkiläinen et al. (2013) report that each additional percentage point of canopy cover within an area decreased the throughfall by 0.05mm. Given the above, the literature would suggest that methods to maximise rainfall interception by a city’s urban forest should focus on increasing canopy area bearing in mind these considerations;

1. Placement of additional trees, 2. Preferential siting of additional trees on impervious surfaces rather than green spaces, 3. Utilisation of wide canopy species, 4. Increasing the proportion of conifers with the urban forest, 5. Ensuring overlap between canopies is limited, 6. Inclusion of understory species

Results: Precipitation in Dublin originates mostly from large storms that produce frequent low

intensity events. Annual rainfall totals about 760 mm and this is distributed evenly throughout the

year (Figure 12). Daily rainfall totals are usually small; since 1942 just 3.5% of days recorded values

of 16 mm or greater. Flooding events then are usually the outcome of persistent events that

saturate the ground over a period of time so that additional rainfall causes water to pool on the

surface (pluvial flooding) and/or rivers to overflow their banks (fluvial flooding). The potential for

flooding is greater in the winter months when evaporation is supressed. Dublin also experiences

coastal flooding as a result of a combination of high tide, low pressure and easterly airflow. Here we

can offer only a crude evaluation of the contribution of trees to water management in Dublin.

Figure 12. Precipitation at Dublin Airport (1942-2015). The table lists the number of days according

to recorded daily precipitation amounts. The graph shows the distribution of monthly rainfall over a

year.

8 Fingal Tree database: http://www.arcgis.com/home/item.html?id=f6a5ffbb0d6d4054b2d83bfc911b73c3

Total (mm)

Frequency (days)

0 11634

0-2 8575

2-4 2737

4-8 2465

8-16 1403

16-32 479

32-64 65

64-128 5

25

Trees can affect pluvial and fluvial flooding by intercepting rainfall and delaying its delivery to

watercourses. Table 16 shows the volume of water intercepted by the tree canopy across Dublin

assuming a uniform annual receipt of 760 mm.

Table 16. The estimated annual water volume (in cubic metres) that precipitates over Dublin

(assuming 760 mm) and the amount intercepted by the tree canopy.

Total volume m3

Intercepted m3

DCC 868,320 90,068

DLR 575,000 94,156

Fingal 893,312 87,507

SDCC 716,448 70,510

Total 3,053,081 342,242

There are a couple of caveats: first, as the tree stock is dominated by deciduous species their

capacity to intercept rainfall is greatest in summer time and will be smaller in winter and; second,

precipitation varies across the study area, being highest in the south, close to the Dublin-Wicklow

mountains. So, a value closer to 200,000 cubic metres might be more reasonable. The value of this

ecosystem service depends on how much of it will eventually make its way into the drainage system

for treatment, and the unit cost of water treatment. Currently, commercial entities pay

approximately €2/m3 for the treatment of wastewater (Walker, 2015). If we take the above

estimated volume of annually intercepted rainwater as an example, and assume 50% of all

intercepted water would require treatment, this would translate to €200, 000 of avoided costs.

These impacts on the city on local hydrology and flooding are strongly linked to the proportion of

impervious surface cover (Figure 13), which is strongly concentrated in Dublin city centre. Much of

the landscape in the case study area is not urbanised and is permeable; planting trees in this

landscape has the same delaying effect however the underlying soil also acts to retain water, much

of which is subsequently returned to the atmosphere via evaporation. In impervious landscapes,

where there is little soil storage potential, the impact of trees could be greater as it can act as

‘umbrella’ slowing the delivery of water to the underlying roads.

26

Figure 13. Proportion of impervious surface within each study area grid cell

27

However, these functions must be placed in the context of the character of flood risk. Figure 14

shows estimated pluvial and fluvial flooding patterns for the Dublin urbanised area; these coverages

are generated by the OPW and represent 1% Annual Exceedance Probability (AEP) or the likelihood

of re-occurrence once in a 100 years. The river patterns show the influence of natural hydrology (as

water accumulates along the water course) and of urbanisation, which has restricted the river

channel and in some cases has placed the river in a culvert. The pluvial patterns are summarised

from the patches of surface water that accumulate in low lying and saturated areas following rainfall

events. For comparison, the tree cover has been superimposed on this map. There is little spatial

correspondence between these coverages, though one site that is characterised by both high flood

risk and very low levels of tree canopy is in south-west Dublin. This area has been identified by SDCC

as highly vulnerable to flooding9 and significant expansion of the urban canopy in this area would be

an appropriate part of flood mitigation measures.

Figure 12. The distribution of trees and predicted flood hazard (medium probability) associated with

rivers and rainfall events. The hazard information is courtesy of the OPW.

9 http://www.southdublindevplan.ie/sites/default/files/documents/SDCC%20SFRA%20Feb%2023%202016.pdf

28

3. Conclusions & Recommendations A tree canopy cover of 15% would make Dublin comparable to other European cities (Table 10);

currently cover across the county is 10% but there are large variations between local authorities.

Even still, the environmental services provided by the existing tree stock are considerable: the value

of intercepted run-off is estimated at €200,000 and that for CO2 sequestered at €622,212 per annum

(assuming a carbon credit value of €6/ton10). The four local authorities are distinct in terms of size,

degree of urbanisation and public/private split (Table 9). Given this variation, no single policy

measure is appropriate for all authorities, or indeed for all areas within each authority. Policy

options for increasing tree canopy cover will be discussed generally, followed by each local authority

in turn.

For rural areas, being largely in private ownership, the policy options are limited to incentivisation.

Currently existing incentivises include the Afforestation Grant and Premium Scheme 2014-202011,

Native Woodland Scheme12 and the FEPS Scheme13 are aimed at increasing forestry cover; the local

authorities could promote these incentives throughout their rural areas. In terms of direct action,

the authorities could ensure that existing canopy, for example present in hedgerows, is retained

through the strict enforcement of hedgerow cutting legislation.

Table 17. The size, proportion of urban land in public ownership and degree of urbanisation for each

local authority.

Total Urban Public Proportion Proportion Urban : Rural

Area (ha) Area (%) Urban Rural Ratio

DCC 11,772.05 55.45 0.97 0.03 32.95

DLR 12,660.04 34.94 0.60 0.40 1.49

Fingal 45,806.36 40.04 0.26 0.74 0.35

SDCC 22,350.01 39.41 0.42 0.58 0.73

Total 92,588.47 41.39 0.43 0.57 0.77

In urban areas, which have a higher proportion of publicly owned land, more options are available

for direct action by the local authorities. Authorities can increase number of trees on their directly

administered land through a range of mechanisms, and indeed this is existing policy in several of the

authorities14. An important consideration for increasing tree canopy in all urban areas is the amount

and distribution of street trees. Streets make up the largest component of publicly administered

land within each of the authorities and, unlike other public lands (e.g. parks) which are unevenly

distributed within urban areas, streets are closely associated with residential populations.

10 https://euobserver.com/environment/132045 11 https://www.teagasc.ie/media/website/crops/forestry/AfforestationSchemeEd2190315.pdf 12http://www.woodlandsofireland.com/native-woodland-scheme-nws 13 http://www.agriculture.gov.ie/press/pressreleases/2007/january/title,13210,en.html 14 See: Halpenny, K. J. (2010) The Forest of Fingal. A Tree Strategy for Fingal. Fingal County Council Parks Division, Community, Recreation Senior Parks Superintendent & Amenities Department; DLR (2011) A Tree Strategy for Dun Laoghaire-Rathdown 2011-2015. Dun Laoghaire-Rathdown County Council; SDCC (2015) Living with Trees. South Dublin County Council’s Draft Tree Management Policy 2015 - 2020. South Dublin County Council. DCC (2016) Dublin Draft Tree Strategy 2016-2020. Dublin City Council.

29

The Dublin City Council (DCC) is dominated by urban land cover (97% urban), with less space for

new tree planting compared to other local authorities. This makes the realisation of large scale

canopy increases a difficult prospect, particularly if this canopy were to be contiguous, as this would

imply the removal of some built land (i.e. buildings, roads, etc.) to make space for trees. The need

for increased canopy cover is clearest for Dublin city centre which has a low cover compared with

other European city centres.

Dublin city council is likely to have an increased opportunity to directly implement such a policy

compared to the other authorities due to the high level of land in public ownership (~55%), due in

large part to the high road density with the authority. Increased street tree planting could and

should be integrated into all road maintenance/traffic management procedures. The total road

length within DCC is 1,550 km, if the average distance between street tree plantings is assumed to

be 20m, this suggests a theoretical maximum of 77,500 street trees could emplaced. In reality the

maximum number will be lower due to context specific factors (e.g. siting conflicts with signage,

utilities and infrastructure), but this theoretical maximum is useful to consider as an upper limit, and

is comparable to the current estimate of 60,000 street and roadside trees present within all of

Dublin county1516.

Dun Laoghaire-Rathdown (DLR) is the “leafiest” authority, in that it has the highest proportion of

canopy cover. This is fortunate, as it has lowest proportion of public land on which it can take direct

action (34.94%) and (unsurprisingly) lowest amount of public canopy (33.02%). DLR has 821 km of

road, suggesting space for a maximum of 41,000 street trees. Ballinteer, Dun Laoghaire, Stillorgan,

Leopardstown and Stepaside are among the areas which should be prioritised for tree planting, as

many parts of these areas have less than 10% canopy cover.

Fingal is the least “leafy” in terms of proportion of canopy cover, though due to its large size it has

the second largest amount of canopy in terms of area. Fingal’s rural areas are used for large scale

and productive agriculture activities, which may make afforestation incentives less effective

compared to SDCC or DLR. Somewhat counterintuitively, Fingal’s continued and fast paced

urbanisation (CSO, 2016) provides the authority with an opportunity to increase canopy by

mandating street tree planting into the design of any new developments. At present Fingal has 1,506

km of road suggesting space for a maximum of 75,328 street trees, which is approximately double

the estimated 33,000 street trees within the Fingal Street Tree database. In terms of areas to be

prioritised, approximately two thirds of urban Fingal has below 10% canopy cover meaning most

areas of Fingal would benefit from additional plantings. That being said, the more northerly towns in

the county, i.e. Lusk, Rush, Donabate, Ballbriggin and Skerries, could be said to be in particular need

of additional tree planting.

South Dublin County Council (SDCC) displays a strong urban/rural divide in terms of canopy cover.

Rural areas contain approximately double the canopy cover per unit area compared to its urban

areas, due in part to the presence of forested areas in the south Dublin Mountains and areas

adjacent to the river Liffey. SDCC’s urban areas are impoverished in terms of tree canopy compared

15 2016. Dublin Draft Tree Strategy 2016-2020. Dublin City Council. 16 2015. Living with Trees. South Dublin County Council’s Draft Tree Management Policy 2015 - 2020. South Dublin County Council

30

to neighbouring DLR, with approximately two thirds of urban SDCC having less than 10% canopy

cover. As with Fingal, most urban areas of SDCC would benefit from increased plantings, though in

contrast to Fingal SDCC has a relative abundance of rural canopy. Areas of particular need of tree

planting are Adamstown, City West, Firhouse, Jobstown, Tallaght, Palmerstown, Perrystown and

Walkinstown. In terms of potential space for street trees, SDCC has 1,084Km of road suggesting

space for a maximum of 54,216 street trees.

4. References Asadian, Y., and M. Weiler. 2009. A new approach in measuring rainfall interception by urban trees

in coastal British Columbia. Water Quality Research Journal of Canada 44. Berland, A. 2012. Long-term urbanization effects on tree canopy cover along an urban–rural

gradient. Urban Ecosystems 15:721-738. Boeur, C. 2010. District of Sechelt Urban Forest Plan. Brennan, M., G. Mills, and T. Ningal. 2015. Assessment of Tree Canopy within the Greater Dublin

Area: Methodology Report School of Geography, University College Dublin. Cappiella, K., T. Schueler, and T. Wright. 2005. Urban Watershed Forestry Manual Part 1: Methods

for Increasing Forest Cover in a Watershed. Casalegno, S. 2011. Urban and peri-urban tree cover in European cities: Current distribution and

future vulnerability under climate change scenarios. INTECH Open Access Publisher. Clayton, C. 2016. The Forest for the Trees: A Comparative Analysis of Urban Forestry Regimes in

Seattle, Washington and Portland, Oregon. University of Washington, Seattle. Coble, D., and M. Walsh. 2012. St. Louis Urban Tree Canopy Assessment. Retrieved October 16:2014. Corine Land Cover. 2000. Corine Land Cover. European Environment Agency, Copenhagen. Crowley, J. W. 2011. City of Redlands Safe Routes to Schools Shadow Mapping. Davies, H., and K. Doick. 2015. Valuing the carbon sequestration and rainwater interception

ecosystem services provided by Britain’s urban trees. NATURE-BASED SOLUTIONS TO CLIMATE CHANGE IN URBAN AREAS AND THEIR RURAL SURROUNDINGS - LINKAGES BETWEEN SCIENCE, POLICY AND PRACTICE, BONN, GERMANY.

Donovan, G. H., and D. T. Butry. 2011. The effect of urban trees on the rental price of single-family homes in Portland, Oregon. Urban Forestry & Urban Greening 10:163-168.

EPA. 2012. CORINE Landcover 2012. Ireland Final Report. Environmental Protection Agency, Johnstown Castle, Co.Wexford, Ireland.

Farrugia, S., M. D. Hudson, and L. McCulloch. 2013. An evaluation of flood control and urban cooling ecosystem services delivered by urban green infrastructure. International Journal of Biodiversity Science, Ecosystem Services & Management 9:136-145.

Fryer, D. 2014. Tree cover in Wales’ towns and cities. Greater London Authority. September 2015. Measuring Tree Canopy Cover in London. London SE1

2AA, UK. Halter, A. D. 2013. Determining existing, possible, and preferable urban tree canopy for Austin,

Texas. Heynen, N., H. A. Perkins, and P. Roy. 2006. The political ecology of uneven urban green space the

impact of political economy on race and ethnicity in producing environmental inequality in Milwaukee. Urban Affairs Review 42:3-25.

Inkiläinen, E. N. M., M. R. McHale, G. B. Blank, A. L. James, and E. Nikinmaa. 2013. The role of the residential urban forest in regulating throughfall: A case study in Raleigh, North Carolina, USA. Landscape and Urban Planning 119:91-103.

Jacobs, B., N. Mikhailovich, and C. Delaney. 2014. Benchmarking Australia's Urban Tree Canopy: An I-Tree Assessment, Final Report.

31

Jim, C., and W. Y. Chen. 2008. Assessing the ecosystem service of air pollutant removal by urban trees in Guangzhou (China). Journal of environmental management 88:665-676.

Jorgensen, Z. 2011. Urban Forest Benefits Report. St. Paul, Minnesota. Lee, J.-H., Y. Ko, and E. G. McPherson. 2016. The feasibility of remotely sensed data to estimate

urban tree dimensions and biomass. Urban Forestry & Urban Greening 16:208-220. Levia, D. F., and E. E. Frost. 2006. Variability of throughfall volume and solute inputs in wooded

ecosystems. Progress in Physical Geograph 30:605–632. Link, T. E., M. Unsworth, and D. Marks. 2004. The dynamics of rainfall interception by a seasonal

temperate rainforest. Agricultural and Forest Meteorology 124 171-191. Llorens, P., and F. Domingo. 2007. Rainfall partitioning by vegetation under Mediterranean

conditions: a review of studies in Europe Journal of Hydrology 335:37-54.

Locke, D. H., and J. M. Grove. 2016. Doing the hard work where it’s easiest? Examining the relationships between urban greening programs and social and ecological characteristics. Applied Spatial Analysis and Policy 9:77-96.

Locke, D. H., J. M. Grove, M. Galvin, J. P. O'Neil-Dunne, and C. Murphy. 2013. Applications of urban tree canopy assessment and prioritization tools: Supporting collaborative decision making to achieve urban sustainability goals. Cities and the Environment (CATE) 6:7.

McGee, J. A., S. D. Day, R. H. Wynne, and M. B. White. 2012. Using geospatial tools to assess the urban tree canopy: Decision support for local governments. Journal of Forestry 110:275-286.

McPhearson, T., D. Maddox, B. Gunther, and D. Bragdon. 2013. Local assessment of New York City: Biodiversity, green space, and ecosystem services. Pages 355-383 Urbanization, biodiversity and ecosystem services: Challenges and opportunities. Springer.

Merry, K., J. Siry, P. Bettinger, and J. Bowker. 2014. Urban tree cover change in Detroit and Atlanta, USA, 1951–2010. Cities 41:123-131.

Mills, G., Anjos M., Brennan M., Williams J., McAleavey C. and Ningal T. (2015) The green ‘signature’ of Irish cities: An examination of the ecosystem services provided by trees using i-Tree Canopy software. Irish Geography, 48(2), 62-77, DOI: 10.2014/igj.v48i2.625

Morani, A., D. Nowak, S. Hirabayashi, G. Guidolotti, M. Medori, V. Muzzini, S. Fares, G. S. Mugnozza, and C. Calfapietra. 2014. Comparing i-Tree modeled ozone deposition with field measurements in a periurban Mediterranean forest. Environmental Pollution 195:202-209.

Nowak, D. J., E. Robert III, A. R. Bodine, E. J. Greenfield, A. Ellis, T. A. Endreny, Y. Yang, T. Zhou, and R. Henry. 2013. Assessing urban forest effects and values: Toronto's urban forest.

Nowak, D. J., R. A. Rowntree, E. G. McPherson, S. M. Sisinni, E. R. Kerkmann, and J. C. Stevens. 1996. Measuring and analyzing urban tree cover. Landscape and Urban Planning 36:49-57.

O’Riain, G., M. Tubridy, and O. Sheridan. 2006. Habitats Survey of High Biodiversity Value Areas in Dublin City, 2006. Dublin City Council and Heritage Council.

Pollefeys, M., R. Koch, and L. Van Gool. 1999. A simple and efficient rectification method for general motion. Pages 496-501 in Computer Vision, 1999. The Proceedings of the Seventh IEEE International Conference on. IEEE.

Potyondy, P. J. 2011. A Comparative Analysis of Urban Tree Canopy Assessment Methods in Minnesota.

Rahman, F., and B. K. Agarwal. Conservation Of Wetlands in India: Rethinking Wetland Laws. Environmental Crisis And Conservation:193.

Richardson, J. J., and L. M. Moskal. 2014. Uncertainty in urban forest canopy assessment: Lessons from Seattle, WA, USA. Urban Forestry & Urban Greening 13:152-157.

Sanders, R. A. 1986. Urban vegetation impacts on the hydrology of Dayton, Ohio. Urban Ecology 9:361-376.

Schwarz, K., M. Fragkias, C. G. Boone, W. Zhou, M. McHale, J. M. Grove, J. O’Neil-Dunne, J. P. McFadden, G. L. Buckley, and D. Childers. 2015. Trees grow on money: Urban tree canopy cover and environmental justice. PloS one 10:e0122051.

32

Shields, E. 2007. CITY OF KIRKLAND. Strohbach, M. W., and D. Haase. 2012. Above-ground carbon storage by urban trees in Leipzig,

Germany: Analysis of patterns in a European city. Landscape and Urban Planning 104:95-104.

Szantoi, Z., F. Escobedo, J. Wagner, J. M. Rodriguez, and S. Smith. 2012. Socioeconomic factors and urban tree cover policies in a subtropical urban forest. GIScience & Remote Sensing 49:428-449.

Tallis, M., G. Taylor, D. Sinnett, and P. Freer-Smith. 2011. Estimating the removal of atmospheric particulate pollution by the urban tree canopy of London, under current and future environments. Landscape and Urban Planning 103:129-138.

Walker, N. 2015. Reforms to water and wastewater tariffs. Ibec policy brief. Walton, J. T., D. J. Nowak, and E. J. Greenfield. 2008. Assessing urban forest canopy cover using

airborne or satellite imagery. Wilson, J. S., and G. H. Lindsey. 2009. Identifying urban neighborhoods for tree canopy restoration

through community participation. Pages 29-42 Planning and Socioeconomic Applications. Springer.

Xiao, Q., and E. G. McPherson. 2002. Rainfall interception by Santa Monica's municipal urban forest. Urban Ecosystems 6:291-302.

Xiao, Q., and E. G. McPherson. 2011. Rainfall interception of three trees in Oakland, California. Urban Ecosystems 14:755-769.

Xiao, Q., E. G. McPherson, J. R. Simpson, and S. L. Ustin. 1998. Rainfall interception by Sacramento's urban forest. Journal of Arboriculture 24:235-244.

33

Appendix 1. Estimating canopy cover: Comparison with other methods

In this section of the report we compare the canopy results of the systematic sample method with

that based on image analysis of a high-resolution IR image, which covers a small section of the study

area. The area was selected to include Phoenix Park and extend from areas outside Dublin into the

city centre.

Figure A1.1. The IR image and its location within the study area.

34

The image itself is taken from the P1 Pleiades17 satellite and provides orthorectified data in four

spectral bands: blue, green and red, which combine to represent visible radiation (VIS) and near

infra-red (NIR). The date of the image was 20/07/2013 and its resolution is (pan/colour) 50cm/2m.

Figure A1.2. The digitised canopy cover using the IR method.

Tree canopy were extracted using ENVI (Environment for Visualizing Images) software which

includes feature extraction in its toolset. In the first step, vegetative cover is extracted by obtaining

the normalised difference vegetative index (NDVI),

𝑁𝐷𝑉𝐼 =𝑁𝐼𝑅 − 𝑉𝐼𝑆

𝑁𝐼𝑅 + 𝑉𝐼𝑆

In the second step, segmentation is used to identify shapes and texture. In this procedure, the user

identifies tree canopies and the software uses this information to train the software to identify

similar features across the entire image. The net effect is the automated generation of tree canopies

based on the training areas; the user can increase the number of training areas to improve the

result. Even so, the automated procedure will misclassify areas of grassland or shrub cover as tree

canopy cover and the result must be manually inspected to be improved. In this work, considerable

effort was expended on improving the tree canopy cover in Phoenix Park.

17 http://www.satimagingcorp.com/satellite-sensors/pleiades-1/

35

Figure A1:3. Examples of digitised canopy cover using the IR method for the Phoenix Park and Dublin

city centre areas.

36

Comparison of canopy cover: The sample method used in this study was compared to the IR method

for the areas where there was sufficient information. The approach taken is to superimpose the

250m grid cell network used in the main body of the study over the canopy cover derived using the

IR method.

1. Sample method: Cell canopy cover was calculated as the percent of sample points that

represented trees and shrubs within each cell. This operation was restricted to those parts

of the image where sampling was done at high density (30 m) and the number of sample

points was ≥20.

2. IR method: Cell canopy cover was calculated by dividing the canopy area in each cell by the

area of that cell.

Figure A1:4. The comparison study area and the cells considered for comparison purposes.

Table A1:1. Canopy cover (%) statistics for the cells depicted in Figure*.

Statistic Sample method

IR method Difference

Mean 11.91 18.56 -0.64

Median 8.00 11.56 0

Standard Deviation 13.84 20.07 9.36

Skewness 1.96 1.40 -0.80

Range 96.00 98.64 94.41

Minimum 0 0 -63.65

Maximum 96 98.64 30.77

Count 2315 2315 2315

The comparative analysis is based on 2315 cells. Overall, the canopy cover for the study area is

higher using the IR method when compared with the Sample method. The distribution of the errors

37

is positively skewed; in other words for any given cell, the IR method returns a higher value than the

point sample method. The differences or errors are well distributed and there are very few cells

where the difference are large >20%

Figure A1.5. The distribution of the error, defined as the difference between the canopy cover (%) calculated by the Sample and IR methods.

Figure A2.6 shows a scattergram of the canopy calculations in which each point represents a cell and

the position represents its value using the Sample (x) and IR (y) methods. The agreement overall is

very good, with a coefficient of agreement (R2) of 78.5% - in other words most of the variation in the

IR estimates can be explained statistically by variation in the Sample estimates.

Figure A1.6. Scattergram of comparison cells based on estimated canopy cover.

38

These differences can also be mapped to see if there are consistent patterns of over(under)

estimation. Overall there is little discernible pattern, which would be indicative of systematic bias.

The greatest discrepancies are found close to the Phoenix park (Sample underestimates) and close

to Ringsend (Sample overestimates).

Figure A1.7. A map of the difference between canopy cover as calculated by the Sample and the IR method. Each cell is 250 m on a side and the comparison is restricted to cells where there are ≥20 Sample points.

Figure A1.8. Detail of a part Dublin where the Sample and IR estimates differ by up to 30%: A) shows

the image; B) the Sample point distribution – yellow dots represent trees/shrubs; C) the IR derived

canopy and; D) the percent differences.

39

Discussion & Conclusion: There are several potential errors that explain the discrepancies between

the two estimation methods. First, the images used by both approaches are different so that trees

may be visible on one but not on another; this is likely a small error as the Sample points generally

coincide with the location of trees on the IR image. Second, the Sample points are incorrectly

classified. Again, this is likely a small error although shrubs and trees can be mixed. In the

comparison, we have added these together but not so in the main report. Third, the IR approach

relies on software to identify features, based on guidance provided by the operator. This is an

iterative process that seems to be effective in identifying vegetated areas (grass, trees, shrubs) using

the NIR band. However, significant errors still remain as the area of tree canopy is generally

overestimated (Figure A1.8) and often includes grass areas and shrubs. This means that intensive

subsequent work is required to ‘clean’ the canopy generated. In this work, the Phoenix Park canopy

was extracted.

Each method has its own advantages and disadvantages. The Sample method:

1. Is time consuming but is easy to operate and is robust as the operator classifies the scene

into clear types; so, it can provide information on building density, paving etc. in addition to

vegetative cover.

2. It provides accurate aerial estimates but the accuracy is dependent on the density of

sampling – greater sampling provides more spatial detail. In this case, we used a systematic

sampling strategy to capture spatial patterns across Dublin.

The IR method:

1. Requires less time to operate but the imagery is expensive to obtain. Image analysis, rather

than generic GIS, software is needed to extract out canopy features but the approach is

effective at extracting vegetated areas.

2. It can provide great spatial detail but considerable work is needed to refine the canopy and

ensure its accuracy.

40

Appendix 2. The databases generated for the DTCS

1. Gridded_Landcover

This Arc shapefile file contains the sample counts for grid cells in Dublin county. The columns refer to the FID (location), total sample (count) and the classification of points into tree, shrub, grass, impervious, road, building, water and barren (bare soil/sand) and area of cell.

FID Count Tree Shrub Grass Imp. Road Build. Water Barren Area

0 6 0 4 0 0 0 0 0 2 159612

1 32 0 28 0 0 0 0 0 4 719710

2 34 0 31 0 0 0 0 0 3 797927

3 0 0 0 0 0 0 0 0 0 24429.5

4 14 0 14 0 0 0 0 0 0 267827

5 41 0 35 0 0 0 0 0 6 980853

6 49 0 48 0 0 0 0 0 1 1000170

7 49 0 49 0 0 0 0 0 0 1000170

8 21 0 20 0 0 0 0 0 1 518675

9 12 0 11 0 0 0 0 0 1 265944

41

2. OPW_Sites

This shapefile consists of the digitised tree canopies in the OPW parks. Digitisation was completed

using Google Earth and Pleiades_RS_Image.

42

3. Pleiades_RS_Image.

This image is a sample of Dublin from the Pleiades satellite. The image resolution is 3m and it

records in the three visible wavelengths and near IR. See appendix 1.

4. Tree canopies

This is a shapefile generated from the Pleiades image which identifies green canopy cover using the

near IR band.

43

5. Google_Image_GDA

This image file was generated from Google Earth and shows the entire study area. Unfortuantely, it

was not possible to extract the canopy cover from the three visible bands in this image.

44

GIS Data Layers

Description Source Type

Gridded Land Cover

This layer contains grids of 1000 m for rural and 250 m for urban with different cover classes and statistics. The attributes include counts of land cover objects in each grid and their proportions in trees, shrubs, grasses, impervious, roads, buildings, water, barren, canopy, green space, impervious and grass.

UCD Geography

ESRI polygon

OPW Sites

This data is on OPW Heritage sites which contains their names and areas. The sites mostly fall in Co. Dublin and a small portion in Counties Fingal and South Dublin.

UCD Geography

ESRI polygon

Pleiades Image

Pleiades RS satellite image with NIR band for vegetation delineation. The image acquisition date is December 2, 2012 and has a 4m resolution. This image was used to extract tree canopies.

Centre National d’Etudes Spatiales (CNES)

TIF

Canopy

This file contains the canopies extracted from the Pleiades satellite image using remote sensing software. The Pleiades image has near infrared band and this band was used along with shape and texture to extract the canopies.

UCD Geography

ESRI polygon

Google Image This is mosaiced Google image covering GDA at 10 m resolution

UCD Geography

TIF

Support Data

Buildings Building footprints described by types and mames OSM ESRI polygon

Land Use Land use descriptions by names and types OSM ESRI polygon

Natural Features

Natural land cover by types and names OSM ESRI polygon

Places Places by types and names OSM ESRI point

Points Points by types and name (Eg. trafic signals, crossing, road junctions, roundabouts, stop signs, gas stations, parking, bus stop, give way signs etc)

OSM ESRI point

Railways Railway type and name OSM ESRI point

Roads Road types, names, speed limit) OSM ESRI Polyline

Waterways Rivers by types, names and widths OSM ESRI Polyline

The datasets used and created in this project (top layer). The support data are generic Open Street

Map files used to help in mapping. All files use the WGS 84 coordinate system apart from

GriddedLandCover, which uses the Irish Grid.